measurements fail on 5.20 beta 18 with Linux

[cgull]

I'm having trouble making measurements on Linux with 5.20 beta 18. This is with the machine's built-in speakers and a USB audio dongle with mic input

When I select Measurement, I get the not-calibrated dialog, select "Continue Anyway". The "Make a measurement" dialog comes up, with its running sound level graph. When I click Start, the graph stops running, an exception is logged, I hear no sweep, and the dialog just sits until I hit Cancel.

Log file and soundcard_debug.txt attached for your viewing pleasure...

I thought I'd seen this when I tried it on my Mac too but when I went back to reproduce it there, it worked fine.

 

[John Mulcahy]

What sample rate are you using? When REW asks the interface for an output channel it is being told there aren't any available that match the requested format. That could be because something else is using the device.

 

[cgull]

48KHz, and since I have no other significant audio apps on the machine, I wouldn't expect the system to be using something other than 44 or 48KHz. Switching to 44KHz didn't help. Uninstalling & installing REW 5.19 did get a successful measurement.

 

[John Mulcahy]

If you select the Force stereo option does it behave any better?

 

[cgull]

No, it doesn't, nor does anything else I've tried. (different drivers, sample rates, etc). I get the same exception every time.

Perhaps worthy of note: checking levels in the Preferences pane works, as does checking levels in the Measurement pane, the signal generator, the SPL meter and the RTA. It really looks like a problem with measurements on Linux, not audio devices.

 

[John Mulcahy]

Could you post a screenshot of the Measure dialog before you press Start?

 

[cgull]

Sorry I disappeared for a while, I got busy with other things. Here's the screenshot with beta 18. Beta 21 behaves the same.

 

[John Mulcahy]

Pretty odd to see the input levels hitting full scale while waiting to measure. You don't appear to have selected an input or an output in the soundcard preferences before grabbing that screenshot.

 

[cgull]

Yes, that was a fresh REW install configured minimally to work at all. Here's screenshots after it was configured to use specific inputs/outputs (mic/speaker on the laptop), with levels adjusted to be somewhat plausible, and showing results from Check Levels. Measurement still fails in this configuration.

regards,

--jh

 

[cgull]

I tried this again twice, first lowering the mic input level and raising the speaker output level to get a better separation between signal and noise, and then muting the mic input completely. Both get the same exception, with the measurement dialog showing "Capturing noise floor..."

regards,

--jh

 

[John Mulcahy]

Could you attach the log file with the exception?

 

[cgull]

Here you go! This is with beta 22, and the same i/o config described in our PMs.

regards,

--jh

 

[John Mulcahy]

The original error relating to not getting any lines is no longer being logged, now the system is logging a null pointer exception for an object which was used without a problem a few lines earlier in the same method and isn't modified in the method. Doesn't make any sense I'm afraid. Perhaps something has failed elsewhere without being logged, but as things stand I have no idea what is going wrong on that system and so no idea how to prevent it.

 

[cgull]

A couple of weeks ago, I tried a clean Ubuntu 19.04 install onto a USB drive, with a beta 22 install. That failed with a similar stack trace. I just came back around to look at this again, and tried beta 24 with both OpenJDK and Oracle JDK, these also failed. Beta 15 with OpenJDK still works for me.

What sort of Linux distro/install are you using for your development/testing? That might be the next thing for me to try.

 

[Antonio Di Motta]

Try checking the mixer setting on linux, because the signal REW is measuring is just noise.
I believe that your sound card has only the mono microphone input and in this hardware configuration you cannot perform the calibration.
In fact the calibration provides two input channels, a reference REF IN and the second measurement IN.
To check if the captured signal is actually noise or signal generated use the RTA and Generator function.
With the Generator function you generate 1Khz sinusoidal and with the RTA function it verifies if you actually have a 1Khz signal level.
If you only see noise you will not find any peak at a high level at 1Khz frequency.

 

[John Mulcahy]

I don't have a Linux test environment, just Windows and macOS - too many flavours of Linux. If you have mono sources best unchecking the "Force stereo" option in the soundcard preferences.

The beta 15 comment is new, previously you said 5.19 worked. Have you tried previous 5.20 beta versions and found beta 15 to be the latest that works on your system?

 

[Antonio Di Motta]

It might be useful to try REW on linux, if you want to limit on the different versions, you could test it right on the
Raspberry platform which in fact almost exclusively uses Raspbian derived directly from Debian.
You would also have the advantage of a very cheap hardware platform and a vast community of software and hardware developers.
Think about it John, my advice is definitely interested but it could be useful to have many others using Linux.
Anyway thanks again for all the support and for creating REW.

 

[cgull]

Try checking the mixer setting on linux, because the signal REW is measuring is just noise.
I believe that your sound card has only the mono microphone input and in this hardware configuration you cannot perform the calibration.
In fact the calibration provides two input channels, a reference REF IN and the second measurement IN.
To check if the captured signal is actually noise or signal generated use the RTA and Generator function.
With the Generator function you generate 1Khz sinusoidal and with the RTA function it verifies if you actually have a 1Khz signal level.
If you only see noise you will not find any peak at a high level at 1Khz frequency.

A bit unclear, but I think you're proposing that perhaps the input REW is reading is not actually hearing REW's generated tones? That's not the case, I check with Preferences/Check Levels, or with Measurement/Check Levels, to see if REW is actually receiving audio from the speakers.

I don't have a Linux test environment, just Windows and macOS - too many flavours of Linux. If you have mono sources best unchecking the "Force stereo" option in the soundcard preferences.

The beta 15 comment is new, previously you said 5.19 worked. Have you tried previous 5.20 beta versions and found beta 15 to be the latest that works on your system?

When I first started with REW a few months ago, beta 15 was the latest available. That did seem to work with the Latitude 3150 under Linux, and later versions do not. 5.19 seems to work also. But I would have to go back and and regression test to find out what the latest working version of REW is there.

It might be useful to try REW on linux, if you want to limit on the different versions, you could test it right on the
Raspberry platform which in fact almost exclusively uses Raspbian derived directly from Debian.
You would also have the advantage of a very cheap hardware platform and a vast community of software and hardware developers.
Think about it John, my advice is definitely interested but it could be useful to have many others using Linux.
Anyway thanks again for all the support and for creating REW.

I think you're speaking to John, but in fact I do have a couple of RPis, I haven't tried them with REW though. What are you using for a microphone/ADC there? And what version of REW?

But today, I discovered that some issue with the front-panel HDMI input on my receiver is causing compatibility problems. After swapping the HDMI cable to a rear input, Windows 10 (1903) on the Latitude 3150 now works and supports 5.1 audio. So I installed REW 5.20 beta24 there.

I have the choice of Java drivers, or ASIO. I'm not sure where ASIO is coming from since I haven't installed ASIO4ALL or any other ASIO driver, my best guess is that it's bundled with the Realtek HD Audio drivers on this machine. ASIO doesn't offer >2 channels on HDMI, and the ASIO control panel is a very simple thing that only sets ASIO sample buffer size.

Sometimes I can get REW to send and receive audio. More often REW can't start the ASIO input, or doesn't receive useful input, and once or twice it has just exited, I forget whether that was from the Preferences dialog or the Measurement dialog.

The configuration that seems to work the best is Java drivers, specifying output device/output on the Denon receiver via HDMI, and input device/input as the Realtek HD ADC and builtin mics. I've "calibrated" that mic to a somewhat random level just to ensure that lack of calibration is not an issue. A Prefs screenshot is below. Attempting to do a measurement shows the same behavior I've described before on Linux, and a familiar looking stacktrace:

Sep 21, 2019 11:06:53 PM roomeqwizard.RoomEQ_Wizard main
INFO: REW V5.20 Beta 24 running JRE 1.8.0_202 64-bit on Windows 10, en, windows-1252
Sep 21, 2019 11:07:03 PM com.synthbot.jasiohost.AsioDriver <clinit>
INFO: Loaded jasiohost64.dll
Exception in thread "Thread-13" java.lang.NullPointerException
        at roomeqwizard.QD.C(y:1060)
        at roomeqwizard.QD.run(y:948)

So I think this issue is not Linux specific, but is instead something about the way I'm using REW. Remember, I don't have a UMIK-1 or any measurement-grade USB mic adapter. I've got an IMM-6 and TRRS extension cable, connected to a laptop's headset jack. For an initial REW smoke test I don't even hook up the external mic, I just use the laptop's internal mic. The laptop is connected via HDMI to a Denon AVR-S740H receiver with a 5.1.2 speaker setup. I don't usually bother to calibrate the mic input level, since I don't have a good reference to check against and have not been too interested in level-dependent measurements (I live in an apartment and so my subwoofer's output limits aren't actually interesting or important). If there's other info that'd be helpful in chasing this down, let me know.

 

[John Mulcahy]

So this is the same laptop but now running Win10? I wish I knew what to ask to understand this, but I have no idea what could be going on. Have you tried any other PC?